b63b60eefc5a39f653577ecd76d817fdb5f4c028
38
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
7470255a22 |
(a1a5d69) Add tags to a ticket with append --tags, keeping the ones it carries, and bump to 0.32.18
`set --tags` writes the whole field, so adding one tag to a ticket that
already carries some meant reading it first and retyping every one of
them.
|
||
|
|
30a2b3429e |
Specify in skill that set overwrites and to use append when that is not desired
|
||
|
|
668263605c | (8a2abf1) Give the wrapper tools for relate, campaign, assign and tags, and bump to 0.32.10 | ||
|
|
e26e5893f7 | (ca2a124) Add --actor-from-git beside every --actor, and bump to 0.32.8 | ||
|
|
b439ad9c41 | (7f772a7) Repair a Status field that lags its directory on mv, and bump to 0.32.6 | ||
|
|
a92dbf552f | Teach the skill to assume WORKLANE_ACTOR and reserve --actor for another's action | ||
|
|
00a64adca8 |
(03bdc76) Draw a symmetric relation once, and refuse the mirror, and bump to 0.32.1
`relate` refused a relation the ticket already held but never looked at what the target held, so both ends of a `relates to` could store a line. `show` then drew the same neighbour twice under the same label, since `RelatesTo` maps to itself in `inverse_label`: one row from the stored line and one derived from the far ticket's mirror of it, indistinguishable. Live on this board, where 8556048 and 8c8a520 each stored the other. Both ends of the fix, because they do different jobs. `relation_rows` drops a derived row that reads exactly like a stored one, which repairs the boards that already hold both directions without an edit. `cmd_relate` refuses to store the mirror in the first place, reporting no change since the edge is already recorded at the far end, which stops it recurring. Neither alone closes it. The dedupe keys on the **label** and the target, not the kind and the target. Two tickets each blocking the other is two facts and the block has to say both, `blocked by` from this end and `blocks` from the other; keying on the kind would fold them into one row and lose half of what the ticket says. Only a symmetric kind can collide with its own stored line, and `RelationKind::is_symmetric` derives that from `label() == inverse_label()` rather than naming `RelatesTo`, so a kind added later answers by construction. The same gate is what keeps the refusal off a cycle: a target holding `blocked by <this>` is not a duplicate. The stored row is the survivor, not the derived one, because it is the line `relate --clear` can address. The consequence is worth knowing and is written down: with both ends holding a line, clearing one leaves the other's, which then draws as the derived row. Correct, since the far end still asserts the relation. 4179958 decided that `relates to` is symmetric and "written once, not onto both files". Nothing enforced it until now. The board's own data is left alone. 8556048 and 8c8a520 still both store the line and the block now draws it once, which is what the criterion asked for. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
0f0767ac00 |
(8556048) Take a relation off with relate --clear, and bump to 0.32.0
`relate` only added. A `--blocked-by` pointed at the wrong ticket was on that ticket forever, refusing every move it made until the target was done or `-f` was passed again, and a target that was deleted did not clean up either: the board carried a dead edge that nothing could clear and every later read of the ticket paid for. `--clear <kind> <target>` takes one off. It is on `relate` rather than on `set` for the reason `RELATIONS` is out of `section_args`: a relation names a target that has to be resolved first, so one writer owns both directions, the way `chown` owns setting an owner and clearing one. It names the target and not just the kind, because a ticket can hold several relations of one kind and emptying all of them is not what anybody is asking for. That is also why it is not spelled the way `set --clear` is, which names a field and empties it whole. Both spellings of the kind are taken, `blocked-by` from the flag and `blocked by` from the label: both are in front of the reader, the flag being what recorded the relation and the label what `show` prints and the file stores, so refusing the printed form would trap anyone clearing a row they had just read. `RelationKind::flag` is a match rather than a `replace` on the label, so the flag, the `--clear` value and the completion list cannot drift. `find_relation` searches the ticket's own edges by id prefix before it searches the board, and that order is the feature rather than an optimisation: no search of the board reaches a target whose ticket is gone, so the stored-target tier is the only thing that can clear a dangling edge, which is the case most likely to need clearing. Two stored targets under one prefix name the candidates instead of guessing, as an ambiguous ticket search already does. An edge the ticket does not hold reports no change, and a target naming nothing at all is an error. The two are different facts: the first is the state the caller asked for, and the second is a typo that a no-op would swallow. The resolver's own reason is carried through, so an ambiguous search says it is ambiguous rather than being reported as nothing found. Recording and clearing the same relation in one call is refused, and the check runs on what was *asked* for and not on what would be written. A relation already in the file is not among the additions, so comparing against those would have let `--blocked-by X --clear blocked-by X` quietly remove it. Removals rewrite the section from what is left and additions are appended to that, in that order and in two passes: a call with no `--clear` then takes exactly the path it always took. The last edge leaves the placeholder, which is the line `append` treats as no content, so the next `relate` replaces it rather than writing underneath it. The two values are one flag, so clap cannot type them apart. `num_args = 2` is per occurrence, which makes an odd number a usage error and every chunk a full pair, and `cleared` parses the kind before the ticket is resolved. Nothing is logged, in either direction, since `relate` never has: a log recording only the destructive half would read as though the additions had never happened. `chowns_clear_is_completed_with_tickets_and_sets_is_not` found the first rule carrying `-l clear` and there are now three, on three commands, each taking something different. It is `each_clear_flag_is_completed_for_its_own_command`, scoped per command, and `relate --clear` offers the kinds rather than a board listing, since tickets where the kind goes would be candidates it refuses. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
7994d98913 |
(76d10d7) Strike through a relation to a discarded ticket, and bump to 0.31.5
The relations block marked a target `○ ◐ ●` or `?`, and none of those can say the target was discarded: the status mark is one cell in a column of five states, and `●` in grey against `●` in green is a colour a pipe does not get. A row pointing at abandoned work therefore read as a live dependency, which is exactly what it is not. The row is struck through instead, the whole of it, gutter and padding included, so the line runs unbroken from the label to the end of the title. The relation stays recorded, for completeness and for the case the target is moved back out of discarded, and reads at a glance as no longer standing for anything. The mark column is untouched: 451fb83 already gave a discarded target the full circle of done in the grey of the untouched pile, and the strike is additive on top of that. `Cell` learns a modifier, the first style on it that is not a colour. Same trick as the palette, one step further: `Modifier::ansi()` is SGR 9 and `From<Modifier> for ratatui::style::Modifier` is `CROSSED_OUT`, so a style that exists for one renderer and not the other is a compile error. Bold stays out of it, because bold rides inside `Color::Magenta.ansi()` and `span_from` adds it back from the colour alone; a second modifier goes beside this one. `paint_cell` opens the modifier ahead of the colour and closes both with one reset, which is why the reset is now a constant the two writers share rather than a literal in `paint`. `span_from` takes the modifier too and has no early return for the uncoloured case any more: a struck plain run is a styled span with no foreground. `Cell::struck` is a builder, because the strike is decided a whole row at a time by the layout that already chose the colours. `Body::push_cell` merges on the whole style rather than the colour alone. Nothing outside the relations block carries a modifier today, and a merge keyed on the colour would silently strike a plain run through the moment one did. The strike is under the same flag every other escape is, so a piped `show` carries none of it. That is the bargain committed and reviewing already take for being told apart by colour alone, and `assets/SKILL.md` says so outright rather than leaving an agent to infer that a plain-looking row may point at abandoned work. Tests are paired both ways, and both halves are load bearing: a renderer that struck nothing fails the presence, and one that struck every row fails the absence. `the_status_marks_are_pinned` had never listed discarded, so it does. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
45b513bdce |
(2733b10) Colour the history block, and indent a comment's body deeper, and bump to 0.31.3
The block shipped with three painted things: the heading, the kind label and the actor with its timestamp as one cyan span. Every value in it was plain, so a status entry drew two status names in no colour while the same names are orange in the metadata table three lines above, and a tags entry drew its tags without the slots every other view gives them. The kind label is now grey, as chrome that names a row. It was orange, which means Status everywhere else, so an `owner` or `tags` label claimed a meaning belonging to the value beside it. `From` on a comment's author line goes grey for the same reason. Each value takes the meaning its field carries in the metadata table: status orange, tags through the same `tag_cells` the Tags field uses, a campaign codename through `tag_color` bare the way `Codename` is, an owner plain, and `_Not set_`, `_None_` and `_missing_` grey. The separator is a constant beside `change`, with `split_change` next to it, so the writer and the reader cannot disagree about what divides the two halves. The actor is drawn through `tag_color`, so one writer is one colour in every entry it wrote, and cyan goes back to meaning the timestamp alone. A comment's body moves to four spaces and the author line stays at column 0, so an indented line is somebody's own words and a line of values at column 0 is the tool reporting itself. The padding is still measured on the unpainted label, now pinned with the alignment test running both ways. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
ab337ca7f8 | (1f1b321) Record a move's reason in its history entry, and bump to 0.31.2 | ||
|
|
5bf0fd1bfe |
(451fb83) Add a discarded status for work that will not be done, and bump to 0.31.0
The flow ended at done, so a ticket that turned out to be a duplicate or obsolete had nowhere to go: it sat in the backlog or was deleted by hand, which destroyed the record and left every relation to it dangling. Discarded is a fifth status with its own directory, keeping the invariant that the filename, the Status field and the directory agree, and keeping done meaning the work landed. Entering it is allowed from any status without -f, because discarding is not a stage of work. Discarded sits last in Status::ALL, which is what makes the rest fall out rather than needing cases: nothing follows it, so a discard clears no stamp and a ticket abandoned from reviewing keeps the journey it made, and every move out is a backward move, so un-discarding needs -f. The two dependency guards now look opposite ways. check_blockers asks what a ticket waits on and gates work moves; check_dependents asks who waits on it and gates the discard, because those plans depended on work about to be declared never-happening. Both skip targets that are themselves terminal. Status::FLOW is the four stages the board draws, so a discarded column on screen stays a136d13's ticket and done stays the last drawn column. Every tree walker keeps using Status::ALL, or a discarded ticket would be unaddressable. util::is_open replaces three spellings of "not done and not discarded", across mv, ls and the campaign standing. The board's blocked mark goes through it as well: that path was only accidentally correct, since a discarded blocker is absent from the FLOW columns, and drawing the column later would have diverged it from what mv allows. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
|
c03f47c52d |
(f774a33) Record a campaign move in the history log, and bump to 0.30.0
assign was the only field writer without an entry: a ticket could leave one campaign for another with nothing recording it. Kind gains Campaign, assign gains the --actor flag the other four writers carry, and the entry is written after the save and behind the no-change return, the rule every writer follows. The entry records Campaign::name, the codename or failing that the title, settled when the entry is written so a rename cannot change what an old entry means. The uuid the field holds says nothing in a log line, and a before-uuid the board no longer has reads _missing_, the way the Campaign field itself renders a dangling one. campaign is the longest mechanical label, so LABEL_WIDTH widens from 6 to 8 and every mechanical row moves two columns right, a deliberate output change that makes this a minor alongside the new flag. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
|
f454d212a4 |
(d4ce964) Record status, owner and tag changes in a history log, and bump to 0.29.0
The comments directory was already an append-only log with one event kind. It is now .worklane/history/<uuid>/, with the kind in the entry and in the filename after the stamp, so a ticket says how it got where it is and not only what was said about it. mv, chown and set --tags each write an entry after their own save and behind their own no-change return, so a refused or no-op edit records nothing. --actor is on all four writers now. An old comments directory is adopted on first access and a file with no marker reads as the comment it is, so nothing is migrated or rewritten. The log moves to src/history.rs, out of util.rs and handlers.rs, and the card's counter counts comments only: a ticket that has moved three times has not been talked about. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
a1bdb5df49 |
(87862df) Attribute an entry to WORKLANE_ACTOR, not to the git user, and bump to 0.28.0
Every comment the CLI wrote was attributed to the git user, so a board several agents write to read as a list of things one human had done. resolve_actor is now the single source: flag, then WORKLANE_ACTOR, then get_git_user(), then unknown. comment --owner is renamed to --actor with no alias, since it names the writer of the entry and not the ticket's assignee, which is chown's. add --owner and campaign add --owner still default to the git user. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
3592584747 |
Edit a campaign's description, and bump to 0.27.0
`campaign set` reached the title and the codename, so the description a campaign was filed with was the description it kept: fixing a typo in the one field nothing could edit meant a hand edit of the file, or refiling the campaign and reassigning every member. Both halves arrive together, which is why d637288 left this out rather than adding the flag. `campaign set --description` replaces the section and `campaign append --description` adds a paragraph to it, because a `set` with no `append` is the half that is easy to reach for and wrong: replacing a description to add a sentence is how the rest of it gets lost. `append` is a campaign's own subcommand rather than the top-level one, which resolves a ticket. `resolve_in`'s tiers are the same over both directories, so one positional naming either object could resolve in each with nothing but a flag to say which was meant, and it would make a liar of the explanation `resolve_ticket` gives a ticket search that names a campaign. The section is the `DESCRIPTION` entry and the surgery is `set_section_body`, `append_section_body` and `Section::normalise`, not a copy scoped to campaigns. `SECTION_ORDER` needs no campaign variant: a campaign holds one section where a ticket holds four, and the followers a campaign does not have are simply not in the file, so its description anchors against `## Metadata`, where `campaign_body` writes it. Still no `--clear`, and that is now a decision rather than a deferral: a campaign's description has no cleared form, so an empty `--description ""` is a paragraph break here as it is everywhere else, and the refusal for an all-blank value says there is nowhere to go back to instead of naming a flag that does not exist. `replaced_items` and `appended_items` hold that check once for both objects, with the way out as the caller's argument. The section refusal no longer says "Ticket", since a campaign reaches it too. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
5dc087d49e |
(9564ac9) Show the annotation in tags ls, and bump to 0.26.4
Since tags.md there are two vocabularies, the one the file declares and the one the tickets carry, and the listing reported only the second. A tag declared and never applied was invisible, and a word somebody invented in a hurry read exactly like one the board had agreed on. Those two cases are the whole reason to declare tags at all. The listing is now the union: still alphabetical, still case sensitive, with the annotation's opening line as a third tab-separated field and a count of zero for a tag nothing carries. **The absence of the third field is what marks a carried but undeclared tag.** Nothing could be appended to the first field without a completion offering a tag that does not exist, and nothing could be added to the second without it stopping being a number. Absence costs neither, and it buys the property a marker could not have: a board with no tags.md has every tag undeclared, so it prints two fields a line, which is byte for byte what this printed before the file existed. An older completion script on a newer binary keeps working, which is the same compatibility 1db02e5 kept for the bare form. A declared tag with nothing said about it answers the placeholder rather than an empty string. An empty third field is a trailing tab and nothing else, so the one difference between declared and not would have been invisible in a terminal and in a completion alike. It is also what the file itself holds under that heading, so the field is the section's opening line either way. `TagIndex::summary` does the cut to one line, the way `CampaignRow::summary` takes one line off a description the read carries whole: the annotation is prose with paragraphs in it, and how much of it a view wants is the view's business. The two fish rules pipe through `cut -f1,3`, since fish takes everything after the *first* tab as one description and the count would otherwise land in the middle of the prose. What is left is the annotation, and a tag the board has never declared is offered with no description at all, which says what the count used to say more weakly. `cut` rather than `string replace` because columns are what is being picked, and `cut -f1` is how a script reads this listing anyway. A test pins both rules rather than the script as a whole, so a rule added later that reads `worklane tags` raw is caught. `assets/ADMIN.md` claimed a declared and unapplied tag does not appear in the listing, which is now the opposite of the truth; it lists at zero. Fixed here rather than left to f7dd1a1, which does not list that file. The same stale sentence in SKILL.md and PLAN.md is recorded on that ticket. Filed on the way past: an annotation carrying a tab splits the listing into a fourth field, 8f650fc, and the `tags set` completion offers the undeclared tags it refuses, which only became filterable now that the listing tells them apart, 193f9be. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
8e2ab6332c |
(8c8a520) Ship a board administration skill, and bump to 0.26.1
Three skills taught the CLI, the review and the grooming sweep. Nothing taught the board itself, and two jobs fell in the gap: deciding what the tags are, now that `tags add` and `tags set` exist to declare them, and putting a board back together when its files and fields disagree. `worklane-admin` is the fourth. It is the one asset written to hold for a board that tracks no software, since a vocabulary skill built around `bug` and `cli` teaches nothing to a board tracking a household, so its worked examples span a tool, a product strategy and a house. Tags as axes rather than as a pile of words, what an annotation has to say to stop drift, which is the boundary against the neighbouring tag and not the definition, and the fact that renaming a tag means rewriting the Tags field of every ticket carrying it. The repairs are by symptom and every one was run against a scratch board rather than reasoned from the source, which is what the ticket demanded and what turned up the one that matters: a ticket whose file sits in `done` while its Status field says `backlog` cannot be repaired by one `mv`, because `mv` reads the current status from the directory and refuses the target it is already in. Two moves fix it. That is 7f772a7. Verifying also turned up that nothing removes a relation, 8556048. The line against `worklane-plan` is drawn in both documents: this one decides the vocabulary, that one applies it across tickets. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
11145b228e |
(c092ea7) Ship the opencode tool definitions with the binary, behind install-tool
`.opencode/tools/worklane.ts` was the second agent asset and the only one that existed nowhere but a checkout. `cargo install` places binaries and nothing else, so anyone who installed worklane got the skills and never the tools, and the copy that did exist was reachable only from this repo. It is now `assets/TOOL.ts`, embedded with `include_str!` like every other asset, and `install-tool` writes it where opencode looks. Its own command rather than a flag on `install-skill`, because the two artifacts are not alike. A skill is inert markdown and one path serves both agents. The tool is executable TypeScript and serves one: Claude Code has no directory it scans for a tool definition, `claude mcp add` is the only way a tool enters its list, and its plugin system adopts agents, hooks, monitors, workflows, output-styles and themes and nothing named tools. Reaching it would mean an MCP server, which is a different artifact and not this file. So somebody on Claude Code takes the documents and never the wrapper, which is the choice this command exists to give them. The path was measured rather than guessed, as the ticket asked. 1.15.10 loads `~/.config/opencode/tool/` and `~/.config/opencode/tools/` both: each was served in turn from a directory outside any project and `/experimental/tool/ids` came back with all eleven worklane tools. Singular is the target, the spelling opencode's own documentation puts first the way it writes `agent(s)` and `skill(s)`. `tools_root()` follows `XDG_CONFIG_HOME` where `skills_root()` does not, because opencode's config root is an XDG path and `~/.claude` is not; `opencode debug paths` reports `$XDG_CONFIG_HOME/opencode` with it set. An empty value counts as unset, the convention `colors_wanted` already follows. `opencode_tool_dir` takes both variables as arguments for the reason `colors_wanted` does: a threaded test binary cannot mutate the process environment to reach one branch at a time. The file name is load bearing and now pinned by a test. opencode builds each tool id as the file name then the export, so `worklane.ts` plus eleven `export const` lines is what makes them `worklane_add` through `worklane_uncheck`. Renaming the file renames every tool and every permission rule naming one. The closing line is not the skills' "on their next session". opencode reads its config once at startup and never reloads it, so a running one keeps the tools it began with and the honest instruction is to restart it. Nothing else has to be installed, which is worth saying because the file is executable and reads as though it wants a toolchain. Criterion 6 asked the command to name the plugin package and Bun as runtime requirements and both are wrong: a copy in a project with no `node_modules` above it registered all eleven tools on a machine with no `bun` on `PATH`. opencode is a Bun binary and the import resolves inside it. The message says the opposite, and the reasoning is on the ticket. `install_each` is the outcome loop both commands now share, with the compare-and-refuse invocation passed in: that line is the only part of an outcome that has to name the command it came from. What each command says after the loop stays with the command. `every_tool_the_wrapper_exports_is_a_real_subcommand` guards the direction that breaks at call time, a tool wrapping a command that has been renamed or dropped. Deliberately not the reverse: the wrapper has no tool for `relate`, `campaign`, `assign` or `tags`, so an agent holding the tools cannot reach campaigns or relations at all, and a test asserting that gap would fail on a decision nobody has taken rather than on a mistake. Its own ticket. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
95426e87be |
(d637288) Guard an existing codename behind --force, and bump to 0.21.1
`campaign set --codename` replaced whatever was there. Once a campaign has a codename it is referred to by it -- on every card in the campaign, at the command line, and in somebody's shell history -- so changing it silently is the mistake `install-skill --force` exists to prevent. So it is refused, and the refusal names both words and the flag rather than leaving it to be found. Checked before any edit is applied, like `SetArgs::contested`, so a call also carrying `--title` writes neither field rather than half of what was asked for. Setting the codename a campaign already carries stays a no-op rather than becoming a refusal: `--force` is about overwriting a *different* word, and there is nothing to overwrite. That is the division `install::Outcome` already draws between `Unchanged` and `Refused`. `--force` guards the codename alone. A campaign always has a title, so guarding that would mean the flag was always required, which is a tax rather than a guard. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
72dbdec2ec |
(d637288) Add campaign set, and bump to 0.21.0
`campaign add --codename` was the only writer of the field, so a campaign already on the board could not be given one without editing the file by hand -- which is every campaign filed before 0.20.0. A `campaign set`, not a flag on something narrower: the file has more than one editable thing, and the moment somebody wants to fix a typo in a title there would be a second command to design. It carries `--title`, which renames the file the way `set --title` does, and `--codename`, through the same parser `campaign add` takes. Built out of `set`'s own parts -- `set_title`, `apply`, `report`, and the whole-file no-op comparison -- so nothing about writing a field lives twice. The codename goes through `set_or_insert_metadata_field` rather than `set_metadata_field`: a campaign filed before the field existed has no such line, and those are exactly the campaigns this is for. That exposed a second thing -- `Codename` was not in `METADATA_ORDER`, so an inserted line landed at the end of the block while `campaign_body` writes it under the ID, two shapes of one file. It is in the list now, directly after ID. `save_ticket` split into `save_entry` with `save_ticket` and `save_campaign` as named wrappers, so a campaign gets the same write-then- rename a ticket has rather than a second copy of it. `write_then_replace` and `find_entry_file` take the noun, so a failure says "campaign" where it is one. No `--description`, filed as 32e98ab: `set` and `append` are a pair for a prose section, and replacing a description with no way to add a paragraph is the half that loses the rest of it. No `--owner` and no `--clear`, each for a reason recorded on the ticket. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
528566d103 |
(61879a4) Surface campaigns on the board, and bump to 0.20.0
A card label and a frame, not a campaign tab. A tab had to say what its columns are, and a campaign groups across the flow rather than within one status, so the answer is the four columns filtered: a filter on the board it already has, not a fifth tab. So a card names its campaign on the label row, before the tags, marked ~ and bold, and the row drops the tags whole before it drops the campaign. `c` opens a picker; taking a campaign filters every column to its members and draws a frame above them carrying the id, the codename, the name, the opening line of the description and the standing. The standing comes from the columns, since while a campaign is selected the columns are its members. `q` takes off one layer at a time: the overlay, the filter, the board. The label needs something to say, which the stored uuid is not, so campaigns gained `Codename`: one word, `v1` or `beaver`, written by `campaign add --codename` and validated at parse time. `campaign ls` prints it and `resolve_in` gained an exact-codename tier above the uuid prefix, or it would be a field you can read and cannot type. Only campaigns carry the field, so the tier is empty for a ticket and ticket resolution is unchanged. The campaigns index is read once per project in `columns_of`, on the pass that already builds the cards, and `digest` now watches the campaigns directory: without it a campaign filed elsewhere never reached the picker. `frame_rows` grew a row, so the two bools deciding the geometry became one `Chrome` and the row indices are named. Left out and filed as d637288: nothing can give a codename to a campaign that is already on the board. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
c65e1b4ed7 |
(55ee6e3) Complete campaigns and tags, and bump to 0.19.0
Every argument naming a campaign or a tag fell through to nothing: `campaign show <TAB>`, `assign <TAB>`, `ls --campaign <TAB>` and `ls --tag <TAB>`. Neither value had a listing a completion could parse. `campaign ls --ids` is the same flag and the same two fields as `ls --ids`, since it answers the same question about a different object. It skips the member counting the decorated listing does, because `members()` reads every ticket on the board and the caller is a shell on a keypress, and it prints nothing on an empty board where the decorated listing prints a sentence. Tags get a command rather than a mode on `ls`. `ls --tag <tag>` answers what carries a tag and cannot answer what tags exist, which is the question a grooming sweep starts from, and a `--tags` beside a `--tag` is one letter from a filter that means something else. `worklane tags` prints the tag, a tab and the count, alphabetical and case sensitive like the filter, so `bug` and `Bug` list separately and land next to each other. assets/PLAN.md used to tell the agent to count the vocabulary with a `show` per ticket; it now points here. The fish half gains `__worklane_nested_arg_of`, which matches a parent and a child, because `campaign show <TAB>` and `chown alice <TAB>` sit at the same token count and the two counting helpers cannot tell them apart. It also gains `__worklane_command_is`, which fixes a leak the campaign and tag rules would have joined. A bare `complete -c worklane -l tag` does not only complete that option's values, it tells fish the whole tool accepts `--tag`: `worklane init --<TAB>` offered `--blocked-by`, `--regression-of`, `--relates-to` and three others. `relate`'s three were already leaking that way, so all of them are scoped now rather than leaving a correctly scoped twin beside a wrong one. `__fish_seen_subcommand_from ls` is not the scope to use, since `campaign ls` is an `ls` fish has seen. The honesty tests grow to five lists and one walker, `every_argument`, since they differ only in the id they look for. Each asserts how many arguments it found, so a walker that stopped finding them fails rather than passing every assertion by never running it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
cc5f9b8a46 |
(74904dc) Let a ticket field be cleared back to unset, and bump to 0.18.0
`set --clear <field>` is the one way to take a value away, repeatable, with the field as a `ValueEnum` value: an unknown one is a usage error listing the six that are not, and the shells complete them for free. Not a per-field `--clear-due`, which is a flag per field to keep in step with the one it undoes, and not an empty value, which is already a paragraph break for a section and would make an accidentally empty variable a silent delete. What goes back is what `add` writes for a field it was never given, through the new `NOT_SET` and `NONE` constants and each section's placeholder, so a cleared ticket and one filed without the field read the same. Sections are in scope and go back through `clear_section_body` rather than `set_section_body`: the latter runs its items through `Section::render`, which would have written `- [ ] _No acceptance criteria provided._`. Owner is `chown --clear <ticket>`, so the field keeps its one writer. The flag carries the ticket because the ticket is chown's second positional and a bare flag would land it in the user's slot for the handler to read back out, which is hand-parsing what clap can state. It writes `_Not set_` rather than the git user: `add` falls back to the git user because it is filing something new, and an owner taken away says nobody owns this. The empty-value error split in two rather than being reworded. `non_empty` says only that the value cannot be empty, for the arguments with no empty form; `clearable` names `--clear`, for the arguments whose field has one. One message for both would have lied about `ls --tag`. Replacing and clearing one field in a single call is refused rather than resolved by whichever was applied last, and cleared fields are reported as `Cleared due on '...'` beside `Updated priority on '...'`, since one list would say `Updated` of a field that was emptied. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
139c9b012f |
(a6023e2) Stamp the time a ticket entered its status, and bump to 0.17.1
mv writes Committed On, Submitted On or Completed On for the status it enters, so the board answers how long a ticket has been where it is without guessing from file mtimes. The three are mutually exclusive: entering stamps, leaving removes and re-entering restamps, so a backward move under -f leaves no stale Completed On behind. These fields are current state, like Status itself; the trail of how the ticket got there belongs to the history log. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
373ba519b3 |
Add a worklane completions command, and bump to 0.17.0
Every argument worklane takes names a ticket, a status, a shell or a skill, and the shell offered a directory listing for all of them. `completions <shell>` prints a script for fish, bash, zsh, elvish and powershell, following what frig does, which is where the shape of this comes from. Two halves, because neither alone is enough. `clap_complete` generates the structure from the clap definition, so a flag added to `command.rs` is completed without anyone touching `completions.rs`, and fish gets a hand-written block appended for what only the board knows. That block runs `worklane ls --ids` in a command substitution, one run per keypress, 2.5ms in release on a board of seventy-five tickets. A status is a `util::Status` value enum now rather than a string, which was the plan's third step and pays for itself three ways. clap refuses an unknown one at parse time naming the four, so `cmd_mv`'s `valid_statuses` and the status half of `cmd_ls`'s error are gone and exit 2 is what an unreal status gets, the bargain `--due` and `--priority` already take. `status_dirs()` and `DONE` are derived from `Status::ALL`, so the flow is written once. And clap emits the four for `mv --status` itself. Only half free: the fish generator emits nothing for the possible values of a *positional*, so `ls <status>` and `completions <shell>` are listed in the appended half after all, read off the enums with `possible_values` rather than by hand. `ls --ids` is the machine-readable listing, and prints the short id, a tab and the title rather than the id alone. Both readers want both fields: fish takes what follows a tab as the description, which is what makes `worklane show <TAB>` a list of titles instead of hex, and a script takes the id back with `cut -f1`. It includes done, where the decorated listing leaves done out, because a completion has to be able to name a finished ticket and `show` is most often pointed at one. It skips a missing status directory rather than refusing, since an error printed over a shell prompt is worse than no candidates, and the listing still refuses one so a person can act on it. The filters all still apply, and an unknown `--campaign` is still refused ahead of everything. Adding the mode tipped `cmd_ls` over the boundary AGENTS.md names, so it takes an `LsArgs`, the way `add` took `AddArgs`. More than the criteria asked, since the cost of each was a name in a list: tickets are completed for `set`, `append`, `relate`, `check`, `uncheck`, the second argument of `chown` and `assign`, and `relate`'s three target flags, plus the skill names read off `install::SKILLS`. The three lists are kept honest by tests that walk the clap definition and fail on a ticket-taking subcommand missing from them, or listed at the wrong position. Nested subcommands are not walked: `campaign` is the only one, none of its three takes a ticket, and the token counting could not tell them apart if one did. Position is counted rather than inferred. `__worklane_first_arg_of` reads `$tokens[2]` and the token count instead of `__fish_seen_subcommand_from`, because `campaign ls` and `campaign show` are also an `ls` and a `show` that fish has seen, and matching against the ticket list to find the position would run worklane again inside the condition. Files are turned off for the command and back on for `tui`, which is the one argument that really is a path. The ticket said worklane takes no file arguments; it takes those. Campaigns and tags are completed nowhere, which needs a listing they do not have: 55ee6e3. And since fish runs whatever is on PATH, the completions are only as current as the installed binary, an older one without `--ids` completing nothing. That is in the README beside the install line. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
6784d80324 |
Let check and uncheck take several criteria, and bump to 0.16.0
Finishing a ticket means ticking every box, and that was one invocation per box: each one resolved the ticket, spliced one line and rewrote the file. The criterion positional is now variadic, so numbers and substrings mix in a single call, and `all` stands for every criterion on the ticket. `all` is a keyword checked before the text, the way the index already is, so a criterion whose own words include "all" is reached by its number or by a longer substring. `is_all_criteria` sits beside `is_index` for that reason. `resolve_criteria` resolves every term before a byte is written and refuses the whole batch on the first term it cannot place, naming it. A call that ticked what it understood and then failed would leave the ticket half-marked with nothing in the file to say which half was meant. The offsets are sorted and deduped, so the report reads in `show`'s order and a criterion named twice, or by both `all` and its number, is marked once. `set_criterion` becomes `set_criteria` and takes every line in one pass. Each named criterion gets a line, printed after the write. One already in the target state reads `Already checked 2: text` and does not count towards the write, so a batch where every box is already ticked writes nothing. The opencode wrapper takes the terms one per line, like the other list arguments. REVIEW.md gains the one thing a reviewer must not do with this: reach for `all`, which would erase the implementer's claim on the criteria the review agreed with. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
ae3ee1bb9e |
Show the target's status beside the relation kind, and bump to 0.15.2
A row now carries two marks, kind then status: what sort of edge this is, then
how far the ticket at the other end has got. `blocked by × ◐ (6ce2c47) Title`.
Two questions are asked of this block, which rows hold this ticket up and
whether the ones that do are finished, and one glyph could not answer both.
○ not started, ◐ under way, ● finished, in grey, orange, cyan and green: the
shape says how far the target has got and the colour tells the two middle
stages apart, since committed and reviewing are both half circles. Green is the
colour the priority ramp deliberately leaves out because it means a title; on a
one-cell mark with no title behind it nothing else wants it, and it is the
obvious colour for done. Orange keeps the meaning it already has, Status.
These are the priority meter's own circles, and what keeps a status from
reading as a level is width: the meter is always two cells, a status mark is
one. the_status_mark_is_one_cell_where_the_meter_is_two pins both halves, and a
card carries the meter and never a status mark while the block carries the
status mark and never a meter.
The kind mark survives a target the board cannot find and the status mark
cannot, since a kind is read from the ticket's own file and a status from the
board. ? holds the column so the id does not move when one target has gone, and
_missing_ is what says it is gone.
Committed and reviewing are told apart by colour alone, so a pipe cannot tell
them apart: both read as under way. Accepted, since ● and ○ are the two answers
usually wanted and show on the target settles the rest, and assets/SKILL.md now
says so outright rather than leaving an agent to work it out from a colourless
half circle.
RelationTarget is back, carrying the title and the status in one Option because
they come from the same lookup. It went out in
|
||
|
|
5ae9c7fe00 |
Mark a relation by its kind rather than its target's status, and bump to 0.15.1
A column of ▁▄▆█ read as undifferentiated blocks. What is worth seeing at a glance is which kind of edge a row is: × for blocked by and blocks, → for relates to, ← for regression of and regressed by. The labels are five long words that blur when one ticket carries several, and the rows holding it up are now the crosses. Both directions of a kind carry one mark, since blocked by and blocks are the same edge read from two ends and the label is what says which end you are on. × is BLOCKED_MARK, the constant a held-up card already draws, in the same brightred. On a card it says this ticket is held up, in the block that this edge is what holds one, which is one word, so one glyph and one definition: the constant moved to util.rs, where both renderers can reach it. Brightred rather than the red asked for, because the id one cell to the right is red and the two would have blended. The arrows are the ones the footer already draws, so they render wherever it does, and no_relation_mark_is_a_priority_glyph pins all three against every level of the meter: the circles belong to priority. The mark now survives a target the board cannot find, because a kind is read from the ticket's own file rather than from the board. Only the name is gone, and `_missing_` in grey is what says so. That is one case fewer than the status mark had, which needed a `?` for a target with no status to show. What this gives up is the status: 6ce2c47's motivation was reading whether a blocker is finished without leaving the overlay, and nothing in the block shows that now. Its criteria 1 and 3 describe the commit before this one, and the ticket says so. RelationTarget went with it. Nothing read the status field once the mark stopped keying on it, and a field written but never shown is a field that goes stale unnoticed, so RelationRow carries the kind and the title and the status is one lookup away on the index if a later ticket wants it. A patch, not a second minor: 0.15.0 was the arrival of a mark and settling which glyph it is belongs to that, where a second minor would read as two features. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
ece201e3ab |
(6ce2c47) Mark each relation with the status of the ticket it names
The Relations block named each target and said nothing about where it was, so the question actually asked of it, whether a blocker is finished, meant closing the overlay and going to find the other ticket. Each row now carries a mark for its target's status, ahead of the id, where it describes the target rather than the row. A rising bar, U+2581 through U+2588 for backlog to done, and ? for a target the board no longer holds. Not a circle: the cards spend those on the priority meter two cells from nothing at all, and another circle would be read as a level. The ramp is also the meaning, so the shape says which status it is. The colour could not. Every colour in the guideline already means one thing and the eight tag slots are reserved, so four status colours could only have come out of something else's meaning. So the shape carries the status and the colour says only that it is one: orange, which is what Status already means, and grey on the missing mark, beside the _missing_ that grey already spells. An unrecognised status reads as missing rather than as a fifth stage, and every_status_has_a_mark pins the pairing against status_dirs so a status added later fails a test instead of quietly reading as ?. The plain text was the real cost. Detail.text was one String drawn by a Paragraph, deliberately uncoloured because an escape sequence there is drawn as literal bytes, which is 6d327df. So the block cannot arrive pre-painted. Resolved.relations is now rows, content_parts returns the body as BodyPart::Text and BodyPart::Relations with the text painted and the rows untouched, and the two renderers take it from there: content_trim paints the rows into escapes, and the board styles the same cells as spans. One builder, two painters. relation_rows decides which edges there are, what they are called and which direction leads; relation_cells decides the layout, the mark and the colours; paint_cells and span_of are the only things that differ. Cell is the unit both take, the same shape the palette uses to stop a colour existing for one renderer and not the other. reference_cells is now the one spelling of a cross reference, shared with the Campaign field. Detail.body is a ratatui Text and is the only thing measured, so last_offset and the Paragraph cannot wrap the body two ways. Detail::plain() flattens it for the suite and is cfg(test): a second representation carried beside the spans is a second thing to disagree. show carries the same mark, decided here rather than discovered later. The rows are one builder so it was free, and two views disagreeing about what a relation shows is a difference nobody can explain afterwards. That is a change to what the CLI prints, so the minor goes on the last commit of this pair. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
c437d87e9e |
(2f3796e) Add a --tag filter to worklane ls, and bump to 0.14.0
Tickets carry a Tags field, but ls could not select on it: finding what is tagged bug meant reading the board. Step 1 of the plan was already done by e41802b, which put tags: Vec<String> on TicketItem through parse_tags, so this is the flag, the filter and the tests. Whole values, not the field. parse_tags has already split and trimmed the comma-separated field, so filter_by_tag compares against those values: --tag bug is not answered by a ticket tagged bugfix, --tag "cli, filter" matches nothing because the field as written is not itself a tag, and a ticket with no tags matches none, however it spells it. Case sensitive, which is the opinion the field already takes everywhere else. tag_color draws Bug and bug in two colours because they are two tags to the field, and a filter that folded them would list tickets carrying a tag nobody asked for while hiding exactly the split a grooming sweep is looking for. One tag per invocation. Several would have to decide between and and or, and that decision is its own ticket, along with substring or regex matching and a tag filter on the board. A tag nothing carries lists nothing rather than being refused, deliberately the opposite of --campaign. A campaign is an object that exists or does not, so it is resolved before anything is read; a tag is free text with no registry, and refusing an unknown one would mean maintaining a vocabulary the format does not have. An emptied status keeps the message ls already gives rather than one per filter naming what excluded them, which is what --priority and --campaign already do. The filter goes through filtered, the one call the three collect sites share, so the counts are filtered by the same set as the rows. Every filter is an Option and none asked for keeps everything, which is what makes them compose in any combination. cmd_ls is now at four filters and the directory, two of them adjacent &Option<String> parameters a caller can swap silently: left as positionals rather than refactored mid-feature, with AGENTS.md recording that the next filter takes an LsArgs struct the way add took AddArgs. assets/PLAN.md said in as many words that there was no tag filter yet, so it would have been actively wrong. It now says what --tag can and cannot answer: it finds what is tagged bug, it does not enumerate the vocabulary, which is the question a grooming sweep starts from. The README's ls row also gained --campaign, which had never been documented there. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
3d6a74bc53 |
(daa0e79) Validate the due date at parse time, and bump to 0.13.3
A clap value_parser beside priority_level rather than a check inside cmd_add, so a bad date is a usage error with nothing written at all rather than written and then rejected. On set --due as well as add --due, from one definition. The planning skill sets dates through set and never through add, so validating only the command that files a ticket would have left the path the validation exists to protect still accepting "next friday". The shape is checked before chrono sees it. %Y-%m-%d accepts an unpadded 2026-8-9, which would be stored as typed and sort before 2026-10-01. Strict on the way in, tolerant on the way out: util::due_date still cuts a datetime down to its date, since a file written earlier can hold one. PLAN.md told the planner the value was not validated. That is corrected, and the guard test that failed on the edit now pins the half of the warning that is still true: a date cannot be cleared, only replaced. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
8d8ae8abfe |
(822e8a1) Add a worklane-plan skill, and bump to 0.13.1
A third asset on the machinery 16c2bed settled, so the wiring is one entry in install::SKILLS. Four jobs, priorities, due dates, tags and campaigns, with propose-then-apply stated once at the top rather than per section. Every job writes to many tickets at once and no field can be cleared back to unset, so the document leans on the two traps that make a sweep destructive rather than merely wrong: set --tags replaces the whole field, and a due date is neither validated nor clearable. feature-planner-agent.md is deliberately not folded in. It decomposes a request into new tickets, which is intake; this orders the tickets that already exist, and one document holding both would fit neither. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
d5fb22cfd8 |
(16c2bed) Add a worklane-review skill, and bump to 0.13.0
install-skill now writes every document the binary carries, with an optional name to narrow it. Requiring a name per document would mean a skill added later never reaches anyone who already knows the invocation, which is the drift the command exists to prevent. --print is the exception and requires the name: a diff is against one document. The review skill is its own document rather than a section of SKILL.md, because the frontmatter description is the whole of what decides loading and one description cannot name both "anything touching a ticket" and "only when a review is asked for". The reviewer never ticks a criterion and never moves the ticket. A tick is the implementer's claim and unticking is the review disagreeing; one mark for both loses the disagreement. mv belongs to whoever does the work. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
5b0bac9c2d |
(7bfb9fc) Add campaigns, a ticket container assigned through worklane assign, and bump to 0.12.0
There was no way to group the tickets that have to be finished together. A goal such as a v1 release was either one ticket that could not describe its parts, or a set of tickets with the goal restated in every description and nothing tying them together. A campaign is a file of its own under `.worklane/campaigns/`, with a title, a description and metadata, and it sits in no status directory. Its state is derived instead: open until every member is done. Deriving it means there is no such thing as a done campaign holding an open ticket, and no second forward-only ruleset beside `mv`'s. An empty campaign is neither open nor done, since calling it done would make a container nothing has been filed against the most complete thing on the board. Membership is one field on the *member*, Campaign in the ticket Metadata, written only by `assign`. The member list is deliberately not on the campaign: that would put membership in two places and make one file that every assignment edits, contended the moment two agents share a board. So `add` gains no `--campaign` and neither does `set`. `add` does not write the field either. Every ticket written before campaigns existed has none, so the insertion path has to exist regardless, and leaving it out of `add` means one path that is always exercised rather than two. That is the Relations argument from 4179958, applied to a field instead of a section. `set_or_insert_metadata_field` is the upsert and `METADATA_ORDER` is its anchor list, the way `SECTION_ORDER` is `ensure_section`'s, so Campaign lands beside Status rather than at the end and Status stays last. A file with no Metadata heading is an error rather than an insertion, per 641e3d5. The three tiers came out of `resolve_ticket` into `resolve_in`, shared by both kinds. A campaign is addressed exactly the way a ticket is, and a second resolver would be a second copy of the precedence rules to keep in step. Two things fell out of that. `Found.dir` is the status for a ticket and empty for a campaign, which is what drops the bracket from the ambiguity listing for one. And `resolve_ticket` can now explain a search that names a campaign, so `mv`, `show`, `chown` and everything else that takes a ticket give the same reason instead of a bare "no ticket found" — one place rather than a check per command. Only on the zero-match path, so an ambiguous ticket error is never replaced by a complaint about a directory the caller never mentioned. The stored value is the full uuid, per the same rule the Relations section follows, and the title is resolved for display. `reference_cell` is now shared between the two, so a dangling campaign and a dangling relation target are both `_missing_` rather than each inventing a spelling. `Resolved` carries what `content_trim` cannot render from the file alone: a struct rather than two more parameters, so the next such field is one entry instead of an edit at every call site. `campaign show` lists done members, unlike `ls`, which leaves the done column out of its listing. What is finished against what is left is the whole question a campaign asks. `ls --campaign` is the other way round and keeps every rule `ls` already has. Deliberately out of scope. The derived progress figure is 1cb999f, now recorded as blocked by this, since nothing here can compute it yet. Unassigning needs a field cleared back to unset, which is 74904dc. And the board gets neither a campaign tab nor a card marker: the detail view resolves the field because it renders the body, and that is all. A marker would want a read per reload and the four corners are already spent, which is the b8ce18c reasoning; filed as 61879a4. Twenty-four integration tests and eleven unit tests, covering assign, reassign, the no-op, a dangling campaign, the ls filter, `mv` refusing a campaign, insertion placement, and a handwritten section and unknown metadata field surviving an assignment verbatim. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
3085b0226f |
(4179958) Add a Relations section for blocked by, regression of and relates to, and bump to 0.11.0
`relate` is the only writer of the section, the way `chown` is the only writer of Owner: a relation names a target that has to be resolved to a uuid first, so `set` and `append` are deliberately kept out of it. One flag per kind, so an unknown kind is a usage error rather than a line refused on the way back out. One direction is stored and the inverse is derived by a scan. `blocked by` goes on the gated ticket, whose move `mv` now refuses while the target is not done, naming every open blocker; `-f` bypasses it beside the backward and skipping moves. A target that is not on the board warns and allows the move. The rendered block is built outside `content_trim` and passed in, since half of it comes from other tickets. The section is created on demand rather than by `add`, so a ticket written before it existed gains it without being regenerated. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
b56b31d2d9 |
(f39b651) Add chown for the ticket owner, and bump to 0.10.0
The Owner field had no writer after `add`. It was folded into `set` in da03690 and comes out here as its own command, so no field is writable from two places, the same reasoning that keeps status in `mv` alone. `worklane chown alice 641e3d5`, user first, after Unix `chown`. That reverses the ticket-first order `mv`, `comment` and `show` use, deliberately: the name borrows the Unix association, so the argument order borrows it too. The swap is left to resolution rather than checked for. Given `chown 641e3d5 alice` the second argument has to resolve as a ticket, and a bare username does not, so the call fails on its own; a check on the first argument would only fire where resolution already refuses and would stay silent on the one case it cannot catch, a ticket whose title is the username. `cmd_chown` is therefore `set_metadata_field` plus `save_ticket`, with the whole-file no-op comparison `set` uses and the same `report()` line, so an owner set to the value it already holds is reported unchanged and nothing is written. Nothing new writes a file: the update goes through `write_then_replace`, scoped to the lines under `## Metadata`, so a description quoting `**Owner**:` is left as prose and a failure leaves the ticket whole. Tested for the change, the no-op, all three resolution tiers, a swapped call, an empty user, a ticket with no Owner field, and that the tick, the handwritten section and the filename all survive. The skill and the opencode tool carry it too, the tool with the argument order in its description, since that is the one place a model can get it wrong without failing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
|
6178d7bd3d |
(dd36e97) Add install-skill, shipping the agent skill in the binary, and bump to 0.8.0
worklane's agent instructions only existed for someone with the repo checked out. Anyone who installed the binary with cargo install got no document at all, because cargo install places binaries and nothing else. The skill now lives at assets/SKILL.md and is embedded with include_str, so an installed copy can never be a different version from the binary that describes it. `worklane install-skill` writes it, `--print` sends it to stdout so a modified copy can be diffed, and a copy that was edited is reported and left alone unless --force: it may have been changed on purpose, and silently discarding that is worse than stopping. One path serves both agents, ~/.claude/skills/worklane/SKILL.md. That was settled empirically rather than from documentation: opencode 1.15.10 loads ~/.claude/skills as an external skill source alongside its own ~/.config/opencode/skill, and with HOME pointed at a temp directory holding only what install-skill wrote, `opencode debug skill` lists the document under the name worklane. Writing opencode's own path as well would be two copies of one document, free to drift, for no gain. The document is new rather than the one under .opencode, which is the user's and not this repo's to ship. The ticket-workflow prose moved out of AGENTS.md into it, so how to use worklane has one writer, and AGENTS.md points there. The mechanism is in install.rs rather than skill.rs, and install() takes its root as an argument: install-tool is coming, it needs a different root and the same compare-and-refuse logic, and passing the root is what lets the tests install into a temp directory instead of the home of whoever runs the suite. Three tests guard the document itself, which is the part that goes stale quietly: the frontmatter has to parse, every `worklane <x>` it shows has to be a real subcommand, and every subcommand outside NOT_FOR_AGENTS has to appear in it, so a new command cannot ship invisible to an agent. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |